The Machine-Intelligence for Networks and Distributed Systems (MINDS) research group develops cutting-edge network systems with a focus on high performance, resilience, and sustainability. By uniquely integrating hardware and software programming, algorithmic theory, and artificial intelligence for system development, we aim to build the infrastructure and logic necessary for deploying next-generation applications.
Two new projects have been funded by Vinnova and Digital Futures!
- SAFIR is a collaborative project between academia (KTH, RISE) and industry (CanaryBit, Ericsson, Nvidia, RedHat, Saab) aiming to build secure, resilient, and trustworthy cloud-edge infrastructures. It focuses on two main goals: (1) enabling trusted execution of AI workloads across cloud and edge environments, and (2) developing secure AI-driven control mechanisms for automated cloud orchestration.
Start/end date: 2025 – 2028
Budget: 27M SEK (2.5M EUR) total; 7M SEK at KTH
Funding body: The Swedish Innovation agency (Vinnova) - EdgeWise is a research initiative in a collaboration with RISE for developing the next generation of trustworthy humanoid robots. The project focuses on improving responsiveness, reliability, and safety through innovative computing and verification techniques that enable robots to operate effectively even under challenging connectivity conditions.
Start/end date: 2025 – 2027.
Budget: 2MSEK (180k EUR) total; 600k SEK at KTH
Funding body: Digital Futures
We look forward to pushing the boundaries of today’s high-speed networking systems in modern datacenters. Welcome, Alexandra! Link to the collaboration announcement between KTH and Nvidia.

From left to right: Dejan Kostić, Marco Chiesa, Changjie WAng, Jonatan Langlet, Alexandra Udrescu, Ali Banaei, Mohammad Siavashi, Farbod Shahinfar. Photo taken by Dejan Kostić.
Our work, accepted at Usenix NSDI 2026, reveals that where you store a packet matters. We show that smart payload placement can drastically cut energy use in terabit-scale NF systems. By tapping into switch buffer queues, we build the first energy-efficient, high-speed packet pipeline—delivering per-flow processing at terabit-per-second rates using just one commodity server and an ASIC switch.
Our project on AI-based code verification had been granted funding from WASP. I am hiring two doctoral students to join this exciting adventure!